Spatially pre-processed target-to-jammer ratio weighted wiener filter using two microphones

In this paper, a speech enhancement algorithm using a two-channel microphone array is presented. The algorithm utilizes the target-to-jammer ratio (TJR) to determine the Wiener filter approximation. The noise in the resulting signal is further reduced, especially under a low SNR situation, using a postfilter-like method. The approximated magnitude-squared coherence (MSC) function for reverberant environments is also used to decide the reduction rate since two microphones provide limited spatial information at low frequencies. The performance of the proposed algorithm is accessed through experiments using real dual-channel recordings. The experimental results demonstrate the usefulness and superiority of the proposed method compared to other generalized sidelobe canceller (GSC) based algorithms using speech quality indices.